Selection question interference item automatic generation method and model based on deep learning

The invention discloses a deep learning-based selection question interference item automatic generation method and model, and the model comprises an answer encoder which is used for encoding an answer of a selection question to obtain an answer representation vector; the article encoder is used for jointly encoding articles and questions to obtain article representation vectors; the context reasoning attention mechanism module is connected to the answer encoder and the article encoder and is used for combining the article context vector and the answer context vector so as to perform context reasoning and generate a final context vector of the encoder; the article context vector is obtained from the article representation vector through an article attention mechanism, and the answer context vector is obtained from the answer representation vector through an answer attention mechanism; and the interference term decoder is connected to the answer encoder to perform decoder initialization by using the output of t